Shapley Value as Principled Metric for Structured Network Pruning

TL;DR

This paper proposes using Shapley values from cooperative game theory as a principled metric for structured network pruning, especially effective in low-data regimes where traditional methods struggle.

Contribution

It introduces Shapley values as a novel, theoretically grounded metric for ranking neural network units during pruning, outperforming existing heuristics in low-data scenarios.

Findings

Shapley values outperform other metrics in low-data pruning.

Random pruning performs comparably with enough fine-tuning.

Shapley values provide a principled approach grounded in game theory.

Abstract

Structured pruning is a well-known technique to reduce the storage size and inference cost of neural networks. The usual pruning pipeline consists of ranking the network internal filters and activations with respect to their contributions to the network performance, removing the units with the lowest contribution, and fine-tuning the network to reduce the harm induced by pruning. Recent results showed that random pruning performs on par with other metrics, given enough fine-tuning resources. In this work, we show that this is not true on a low-data regime when fine-tuning is either not possible or not effective. In this case, reducing the harm caused by pruning becomes crucial to retain the performance of the network. First, we analyze the problem of estimating the contribution of hidden units with tools suggested by cooperative game theory and propose Shapley values as a principled ranking metric for this task. We compare with several alternatives proposed in the literature and discuss how Shapley values are theoretically preferable. Finally, we compare all ranking metrics on the challenging scenario of low-data pruning, where we demonstrate how Shapley values outperform other heuristics.